Why most published congressional trading returns are wrong
A lot of numbers get published about how well Congress trades, and they frequently disagree with each other by a wide margin about the same politician over the same period.
That is not necessarily anyone being dishonest. Congressional disclosures are awkward data to score fairly, and four small methodological choices produce very different headline figures. Understanding which four lets you read any of these numbers — including ours — with the right amount of scepticism.
1. The execution-price problem
Filings disclose a transaction date and a dollar range. They never disclose an execution price. Nobody publishing a congressional return figure, this site included, knows what a member actually paid.
The only defensible consistent choice is the closing price on the disclosed transaction date, which is what we use throughout.
Some published figures instead use the price on the filing date, up to 45 days later. That produces a meaningfully different — and usually more flattering — number, because it silently skips the entire period between the trade and its disclosure. A member who bought before a rally gets credited with the rally twice: once in the entry price, once in the return.
Ask of any figure: priced from which date?
2. Cherry-picked horizons
"Up 40%" means nothing without its window and its end date. The same trade looks brilliant measured to a market peak and mediocre measured a month later, and nothing about the trade changed.
Composite deliberately avoids a fixed horizon. Each buy is scored at its realised return if it was sold, or marked to today's price if still held. There is no chosen measurement date to select favourably, because there is no chosen measurement date at all.
The separate 1W/1M/3M checkpoint figures exist so several honest windows sit side by side rather than whichever one flatters. They tell a consistent story here: +0.29%, +1.60%, +5.15% raw. No window is being hidden.
3. Money-weighted versus pick-weighted
Should a $2,000 trade count the same as a $2,000,000 one?
Reasonable methodologies disagree, and they are answering different questions. Weighting by value answers how did their money do — but filings only disclose bands, so any money-weighted figure is estimating from a midpoint before it starts. Weighting each buy equally answers was each decision a good one, and sidesteps the estimation entirely at the cost of not reflecting portfolio-level returns.
Composite weights every buy equally. Our performance charts, by contrast, are size-weighted — which is why the two can disagree about the same member, and why we say so on the page rather than letting a reader assume they measure the same thing.
Neither is more correct. Conflating them is a common source of published disagreement.
4. Raw versus market-relative return
This is the single biggest source of divergence, and the one most often skipped.
A politician who bought only technology during a technology rally shows a large raw return almost regardless of skill. A raw-return ranking makes them look exceptional. Establishing whether they actually beat the market requires subtracting what the relevant benchmark did over the same window — which is more work, needs a sector classification for every stock, and usually shrinks the number.
How much does it matter? On our data, it is the difference between a strategy and nothing at all: +5.15% raw at three months versus −0.32% measured against each stock's own sector benchmark. The full analysis.
Any figure that does not say what it was measured against has skipped the step that decides the answer.
The four questions
Of any congressional-trading return figure, including every one on this site:
- Priced against what date — transaction or filing?
- Measured over what window, ending when?
- Weighted how — per trade, or per dollar?
- Against what benchmark — nothing, an index, or the stock's own sector?
Different honest answers to those four can move the same trade's "return" by tens of percentage points without anybody lying.
Why we publish our own edges
A site whose premise is measurement accuracy should be the most transparent about where its own measurement has limits, not the least.
So: our sector assignments are published per stock. Our scoring rules are written out in full. The per-trade returns and benchmark comparisons behind every aggregate are downloadable as CSV, so if you think our benchmarking is wrong you can recompute it rather than argue about it.
That is also the fairest test to apply to anyone else's numbers. If a published figure cannot be reconstructed from disclosed data and a stated method, it is not a measurement — it is a claim.